Among the methods for determining the authenticity of the statements, polygraphs are traditionally used, and lie detection using EEG has been attempted in recent years. This method mainly uses the ERP component called P300, which appears after about 3 ...
Among the methods for determining the authenticity of the statements, polygraphs are traditionally used, and lie detection using EEG has been attempted in recent years. This method mainly uses the ERP component called P300, which appears after about 300 to 500 ms in the parietal lobe after presenting the stimulus. The polygraph and P300-based EEG have a high accuracy but are vulnerable to a countermeasure that can intentionally ruin the results. As an alternative to this, an inspection method using an implicit association is promising. The representative implicit measurement method is the autobiographical Implicit Association Test (aIAT). But aIAT attempted so far has a disadvantage in that it can distort the result by analyzing the reaction rate of the testee. In order to prevent this distortion, we tried to classify the truth of the statement using the aIAT method to measure EEG.
We also tried to develop a new implicit method using picture stimuli. Based on the assumption that the subject 's prior experience has an influence on the perception of the object, we developed a photographic test that presents the image stimulus with noises. And confirmed the object' s recognition after mock crime. For example, when a guilty mission is conducted and a guilty subject observes an object, a visual stimulus involving noise will notice the picture faster than the innocent subject, and another form of EEG will appear.
In addition, 'virtual reality mock crime' was constructed to overcome limitations of mock crime experiment configuration for current lie detection research, and 'actual mock crime' was compared. Because of ethical and cost issues, Most of the studies on lie detection are common to conduct the test after 'mock theft'. Our researchers tried to find out whether virtual reality can be used for mock crime by recreating the virtual reality space by taking the place where 'actual mock crime' was made and reconstructing it in virtual reality.
In the first year, we developed the aIAT using EEG and the photo-recognition test through preliminary experiments. In main experiment, subjects conducted the guilty mission or the innocence mission, and we examined whether each test with EEG correctly discriminated groups. As a result, there was a significant difference between the groups in the reaction time, which is the time required to press the button, and the D-score was 73.3% accuracy to distinguish groups. In the EEG, ERP amplitude differecen between guilty and innocent from 600 and 800 ms after presentation of stimuli in the anterior frontal area of the head was statistically significant. EEG distinguish the groups 80% accuracy.
In the photo-recognition test, when the guilt-related objects (guilty objects) were viewed, the guilty group recognized the objects at a faster rate than the innocent group. In spite of the noises, statistical differences between the groups were significant. There were also a significant power differences in EEG in relative low frequency bands (Delta, Theta, Alpha), but it was not enough to make individual diagnosis correctly.
In the second year, 3D camera was used to shoot the mock crime scene and imitated as 3D virtual reality (Virtual reality mock crime). Using the most widely used P300 - Guilty Knowledge Test (GKT), the GKT test results by subjects conducted ‘actual’ mock crime were compared with the results by ‘virtual’ mock crime subjects. As a result, the group conducted mock crime with a virtual reality showed almost the same EEG waveform patterns as the actual mock crime group. And it was revealed almost the same probability in individual discrimination. This shows that virtual reality mock crime can replace actual mock crime when P300 based GKT is used as a test method to lie detection.